An empirical dataset combining repository metrics and pipeline workflows for predictive software analytics

This data article presents a comprehensive dataset that captures the workflow history of Continuous Integration and Continuous Deployment (CI/CD) pipelines, along with the code-level repository metrics. There are 303,079 distinct execution records from August 2016 to April 2026 in the dataset. Specifically, the data comprises of archival Travis CI records from August to December 2016 which serves as the baseline, and live GitHub Actions workflows from May 2024 to April 2026. The data collection used a custom python extraction script which combined historical Travis CI data and live GitHub Actions workflow metadata with deep Git commit history. The chosen live repositories covered 34 highly active open-source projects across different domains like DevOps tools such as kubernetes/kubernetes, web frameworks like facebook/react, and machine learning like tensorflow/tensorflow. The variables collected consist of pipeline execution durations, pipeline final status, the conditions that triggered the events, code churn (additions and deletions), and the number of changed files. This resulted in a structured and tabular format (csv) dataset which is hosted and accessible on Mendeley Data and serves as a solid foundation for researchers and practitioners interested in discovering patterns in build breakages, creating predictive machine learning models for forecasting CI/CD pipelines' workflows, and advancing AIOps methodologies in software engineering.